Alibaba will ban employees from using Anthropic's artificial intelligence tools for work beginning July 10, citing back-door security risks from the U.S. company, CNBC confirmed Monday. The move escalates a direct confrontation between the Chinese e-commerce giant and the maker of Claude, after Anthropic accused Alibaba of a "distillation attack" designed to extract proprietary AI capabilities.
Anthropic's Claude Code has been placed on a high-risk software list, according to people familiar with the internal policy who asked not to be named. Alibaba employees must uninstall all Anthropic models and agent products and switch to the company's own AI assistant, Qoder, by the Friday deadline.
Ban follows 'distillation attack' accusation
Alibaba's decision comes weeks after Anthropic sent a letter to the U.S. Senate Committee on Banking, Housing, and Urban Affairs, blaming the Chinese tech company for what it called "the largest known distillation attack" on Anthropic to date. The letter accused Alibaba of "brazenly" and "illicitly" attempting to extract its AI capabilities.
Anthropic's terms of service prohibit Chinese companies and those from other "adversarial nations" from using its models. The ban at Alibaba aligns with a wider online backlash in China, where posts on Reddit and GitHub have outlined hidden code meant to detect users who might be based in the country.
Loopholes and workarounds across the industry
The Financial Times reported Friday that Anthropic is moving to close loopholes that let Chinese firms bypass restrictions and access Claude through third countries. The newspaper said Chinese fintech group Ant "had provided employees with corporate Claude accounts that were accessed through the company's intranet, which is connected to its Singapore-based entity." Ant declined to comment on the FT report.
ByteDance, the parent company of TikTok, does not facilitate direct access to Claude but started a reimbursement program on April 2 that allows engineers to expense personal subscriptions, a person familiar with the matter told CNBC. Staff can access those subscriptions on virtual private networks. ByteDance declined to comment. The policy is meant to encourage employees to "experience and learn" about a broader range of AI products, the person said.
Why this matters for IT and development professionals
For IT and development teams, the episode reveals how geopolitical tensions can abruptly disrupt access to tools already woven into daily workflows. As AI for IT & Development becomes central to coding and infrastructure tasks, vendor restrictions and national security policies create compliance risks that many organizations have not fully planned for. Choosing a coding assistant or AI platform now requires assessing not just performance but also long-term access guarantees across jurisdictions.
Your membership also unlocks:
